A first draft is not a finished thing. This last lesson covers the mile most people skip: editing the output, putting your own voice back in, and making sure nothing generic or wrong reaches your reader.
By the end, you'll have:
Everything up to now was about getting a good draft. This is about making it yours.
The most common mistake with AI isn't a bad prompt — it's sending the first output. It's usually 70% there: roughly the right shape, slightly off in tone, occasionally wrong on a detail, and unmistakably not you.
Treat it like a first draft from a competent colleague. You wouldn't forward that untouched either.
Once you know these, you'll spot them instantly — in AI output and in your own edits:
Everything in threes
"Fast, reliable, and scalable." Lists of three everywhere is a rhythm tell. Break it.
Empty enthusiasm
"In today's fast-paced world…" Filler that sounds positive and says nothing.
"Not just X, but Y"
A construction so common it reads as a fingerprint.
False balance
Every point hedged with "however" and "on the other hand" until it commits to nothing.
Over-explaining
Restating the question, summarising what it just said, telling you what it's about to tell you.
Uniform sentence length
Real writing has rhythm — long, then short. Flat rhythm reads as machine.
None of these are wrong exactly. They're just generic — and generic is what makes readers skim.
Don't try to fix everything at once. Two focused passes works far better:
Pass 1 — substance
Is it true, complete, and about the right thing? Fix facts, missing points, and anything that misrepresents you.
Ask: "Would I put my name on this claim?"
Pass 2 — voice
Now make it sound like you. Cut the filler, vary the rhythm, replace stiff phrasing with how you'd actually say it.
Ask: "Would I say this out loud to a colleague?"
Two moves that work well together — one done by you, one done by AI:
Rewrite the opening and closing yourself. First and last lines carry most of the personality. Write those in your own words.
Feed it your voice. "Here's how I usually write. Rewrite this in that style." (Straight from Lesson 7.)
Thirty seconds. Five questions. Then ship it.
Are the facts right? Names, numbers, dates, claims — spot-check the ones that matter.
Does it sound like a person? Cut any sentence that could belong to anyone.
Is anything private still in it? Names, figures, details that shouldn't travel.
Does it say what you meant? Not what it's about — what you wanted it to do.
Would I sign this? If yes, stop editing. Good enough shipped beats perfect unshipped.
There's no universal rule, but a sensible line: disclose when the reader would be surprised or misled if they found out.
Usually fine without saying
Personal emails, notes to yourself, brainstorming, drafts you've substantively edited and own.
Disclose or avoid
Academic work, journalism, anything presented as your first-hand expertise, and settings with their own rules (school, employer, client).
A great manager doesn't send a junior's draft straight to the client. They read it, adjust it, add what only they know, and put their name on it.
The draft was never the deliverable. The deliverable is what you sign. Edit like it's yours, because it is.
Your feedback shapes what we build next.
You can start a conversation, write a clear brief, pick the right tool, build a habit, verify the work, run a project, steer it with examples, protect your data, and finish what you produce.
That's more than most people using AI daily can claim. From here it's all depth — pick the branch that interests you most.
Now that you can use AI well, understand how it actually works — in plain English.
For engineers and the curious: machine learning, neural networks, NLP, agents, and more.
Worth re-reading before you paste anything into a new tool.
Back to the full tutorial library.
